Provider does not track active state of network

[agorgl]

When creating a libvirt network without autostart and rebooting the hypervisor, the network and the instances (domains) are inactive / stopped, as expected.

Running terraform plan in this scenario we get this:

Terraform used the selected providers to generate the following execution plan. Resource actions are indicated with the following symbols:
  ~ update in-place

Terraform will perform the following actions:

  # libvirt_domain.instance will be updated in-place
  ~ resource "libvirt_domain" "instance" {
        id          = "7dbfcef6-0dd3-4ee2-854e-0f31d3c30083"
        name        = "instance"
      ~ running     = false -> true
        # (10 unchanged attributes hidden)


        # (3 unchanged blocks hidden)
    }

Plan: 0 to add, 1 to change, 0 to destroy.

Changes to Outputs:
  - instance_ip = "10.3.0.61" -> null

Do you want to perform these actions?
  Terraform will perform the actions described above.
  Only 'yes' will be accepted to approve.

  Enter a value: yes

libvirt_domain.instance: Modifying... [id=7dbfcef6-0dd3-4ee2-854e-0f31d3c30083]
╷
│ Error: error creating libvirt domain: Requested operation is not valid: network 'network' is not active
│
│   with libvirt_domain.instance,
│   on main.tf line 80, in resource "libvirt_domain" "instance":
│   80: resource "libvirt_domain" "instance" {
│
╵

This shows that provider is able to track if the libvirt_domain is running and try to bring it up again, but it does NOT! track the state of the network it is connected to and we manually need to run virsh net-start network before running terraform plan again in order to get a successful result.

[dmacvicar]

// resourceLibvirtNetworkUpdate updates dynamically some attributes in the network
func resourceLibvirtNetworkUpdate(d *schema.ResourceData, meta interface{}) error {
	// check the list of things that can be changed dynamically
	// in https://wiki.libvirt.org/page/Networking#virsh_net-update
	virConn := meta.(*Client).libvirt
	if virConn == nil {
		return fmt.Errorf(LibVirtConIsNil)
	}

	network, err := virConn.NetworkLookupByUUID(parseUUID(d.Id()))
	if err != nil {
		return fmt.Errorf("can't retrieve network with ID '%s' during update: %s", d.Id(), err)
	}

	d.Partial(true)

	activeInt, err := virConn.NetworkIsActive(network)
	if err != nil {
		return fmt.Errorf("error when getting network %s status during update: %s", network.Name, err)
	}

	active := activeInt == 1
	if !active {
		log.Printf("[DEBUG] Activating network %s", network.Name)
		if err := virConn.NetworkCreate(network); err != nil {
			return fmt.Errorf("error when activating network %s during update: %s", network.Name, err)
		}
	}

in resourceLibvirtNetworkUpdate, which means resourceLibvirtNetworkUpdate is not called if the network is not update. Because we don't track the active property in resourceLibvirtNetworkRead.

But I am not sure what is the best way to track this. A computed attribute? Tell somehow terraform to foce the update even if no parameter from the schema is changed?

[scabala]

I think this is nice feature to have.

[dmacvicar]

I think the overall network provider is messy because it deviated a lot from libvirt, due to its origin in supporting specific use cases for Kubernetes infrastructure.

I don't have an opinion on what is the best way. I think the best is to continue the v2 network resource that is 1:1 with libvirt (there is a branch for it).